International Research Journal of Advanced Engineering and Science

ISSN (Online): 2455-9024

Implementation of a Kalman-Bucy Filter for

Estimating Product Concentration of a CSTR Process
Eze Paulinus C.1, Muoghalu Chidiebere2, Jonathan Amabikutol E.3,
Aigbodioh Ferdinard A.4, Okoronkwo Emmanuel A.3, Ekengwu Bonaventure O.2
1
Electrical and Electronic Engineering, Covenant Polytechnic, Aba, Abia State, Nigeria
2
Electrical and Electronic Engineering, Chukwuemeka Odumegwu Ojukwu University, Uli, Anambra State, Nigeria
3
Electrical and Electronic Engineering, Federal University of Technology, Owerri, Imo State, Nigeria
4
Electrical and Electronic Engineering, Auchi Polytechnic, Auchi, Edo State, Nigeria

Abstract—This paper has presented the implementation of a Kalman-
Bucy filter for estimating product concentration of a continuous
stirred tank reactor (CSTR). An ideal operation was assumed such
that the chemical reaction taking place in the system was considered
a linear continuous time process. A transfer function model for the
product concentration considering reactor mass balance is obtained.
It is required to estimate and track the product concentration using
measurement noise. A Kalman-Bucy filter was developed and
implemented in Matlab/Simulink environment. A computer
programme using Matlab code for the Kalman-Bucy filter
implemented in Matlab/Simulink embedded block function was used
for the simulation in this paper. The results obtained showed that the
performance of the filter in rejecting the noise and estimating a
signal that accurately tracks the input improved as the value of the
noise variance was tuned within the range of 0.0001 to 0.1.
Keywords— Computer programme, CSTR, Embedded block function,
Kalman-Bucy filter, Matlab/Simulink, Product concentration.
I. INTRODUCTION
Some of the primary factors driving the growth of continuous
stirred tank reactor (CSTR) are increasing adoption of
sustainable production technique to gain competitive
advantage, need of cost cutting and mass production [1]. A
CSRT is a system in which reactants are added and products
are continuously removed at the same rate with the reactant
continuously stirred using internal component [1]. The flow
rate, reaction time and dilution rate are basic three parameters
that define the nature of CSRT. It has gain wide application in
chemical and waste water treatment industries basically due to
the fact that it has approximate mixing property. For instance,
in chemical industry, it provides perfect mixing of chemicals
which are continuously added in the reactor and also gives
similar composition of input chemical and output mixture [1].
It is one of the most essential unit operation in chemical
industries. Continuous stirred tank reactor (CSTR) shows
nonlinear behaviour and its operating is normally wide [2].
The chemical reactions that take place in CSTR are either
endothermic or exothermic. During this reaction, energy can
either be added or removed so as to keep a constant
temperature. It is usually operated in such a way that the well
mixed chemical process normally run at steady state.
There are several works in literature that deal with CSTR.
Mohd et al. [4] carried-out experimental study on the effect of
operating conditions on CSTR performance. The influence of
operating conditions on the conversion and specific rate
constant was studied. Igbokwe et al. [5] presented the
characterization of a five-litre continuous stirred tank reactor.
Antonelli and Astolfli [6] designed bounded control laws for
the temperature stabilization of a class of continuous stirred
tank reactors with exothermic or endothermic reactions using
methodologies and tools from Lyapunov theory. Controller
design for continuous stirred tank reactor is presented by
Prabhu and Bhaskaran [7]. It studied the problem of
temperature control of CSRT using adaptive controller. DI
Ciccio et al. [8] proposed a novel digital control law for
continuous stirred reactor tank. It used the relative degree
preservation under sampling as its methodology. Alejandro
(2010) presented a problem based learning approach in the
analysis of continuous stirred tank chemical reactors with a
process control approach.
In this paper, a Kalman-Bucy filter is developed and
implemented in Matlab/Simulink to estimate the product
concentration of component X in a reactor. The Kalman-Bucy
filter is a Kalman filter in continuous time. The design of
Kalman-Bucy filter is such that it can be used to estimate
unmeasured states of a continuous process, either for
controlling one or more of them [3].
II. THEORETICAL BACKGROUND
This section presents the theoretical frame work for
developing the dynamic model of the considered continuous
stirred tank reactor (CSTR). The state space modelling and
representation of the continuous time process. Also in this
section, the theoretical concept of a Kalman-Bucy filter is
presented.
A. Process Description for Ideal CSTR
In this paper, the reactor mass balance dynamic equation
has been considered for estimating the product concentration
of a component X in a typical reactor. A model of a CSTR
using first principles operational data (Table 1) as stated in
[10] has been used for carrying out computer simulation in
this work. Fig. 1 shows an irreversible, exothermic chemical
reaction taking place in constant volume reactor that is cooled
by a single coolant stream [2]. A feed material whose
composition is C X 0 enters the reactor at a temperature, Tc 0
at a constant volumetric flow rate Q . Product is taken out of
the system at the same volumetric flow rate Q . A

97

Eze Paulinus C., Muoghalu Chidiebere, Jonathan Amabikutol E., Aigbodioh Ferdinard A., Okoronkwo Emmanuel A., and Ekengwu
Bonaventure O., “Implementation of a Kalman-Bucy filter for estimating product concentration of a CSTR process,” International Research
Journal of Advanced Engineering and Science, Volume 3, Issue 1, pp. 97-100, 2018.
 International Research Journal of Advanced Engineering and Science
ISSN (Online): 2455-9024

homogeneous mixture is assumed for the liquid content within TABLE 1. Steady state operating data [10].
the reactor. Serial No. Physical Parameters Symbols Values
1 Reaction rate constant ko 7.2 1010 min 1
2 Feed flow rate Q 100 l/min
3 Reactor temperature T 441.2 K
4 CSTR volume V 100 l
5 Activation energy term E/R 1  104 K

Substituting the values of the parameters in Table 1 into

Eq. (3) gives the process transfer function as:
1
G p ( s)  (4)
s 1
C. The Kalman-Bucy Filter
Given the inputs and measured output, assumptions are
made on the state and output noise. The Kalman-Bucy filter is
used to estimate unmeasured states (assuming that they are
Fig. 1. A typical diagram of a continuous stirred tank reactor.
observable) and the actual outputs of the process. Fig. 2 shows
Fig. 1 shows a jacketed continuous stirred tank reactor the estimated states x̂ , and the estimated measured output ŷ .
(CSTR) in which heat is added or removed as a result of the
difference between the jacket fluid and the reactor fluid [2].
Heat transfer liquid is usually pumped through a nozzle that
circulates the liquid through the jacket at a high velocity. The
coolant flows at a rate, Qc and at a feed temperature, Tc 0 . The Fig. 2. Block diagram of input-output relation of the Kalman-Bucy filter.
coolant liquid temperature at exit is Tc .
A differential Riccati equation to be integrated through
B. Dynamic Model of reactor Mass Balance time is used by the Kalman-Bucy filter unlike the Kalman
In order to perform the modelling of the system, filter that uses a predictor-corrector algorithm to update the
assumptions are made so as to obtain a simplified model for an state estimates. The filter update equations are stated as:
ideal CSTR as follows [2]: K  PC T R 1 (5)
xˆ  Axˆt   Bu (t   G yt   Cxˆt 
i. The mixing taking place in the reactor and jacket is
(6)
perfect
ii. Reactor and jacket volume is constant P  AP  PAT  KRK T  Q (7)
The dynamic modelling for the process is formulated In the above equations, K is Kalman-Bucy filter observer
considering mas balance as in [2]: gain matrix which makes the observer sensitive to sensor
 QC X 0  C X   VrX
dC X noise, P is an estimate of the covariance of the measurement
V (1)
dt error and satisfies the Riccati equation, C T is the transpose of
where C X = product concentration of component X in the the measurement matrix C , R is a weighting matrix of
reactor, rX = rate of reaction per volume, and V = volume of measurement (sensor) noise, Q is a weighting matrix of
CSTR process (state) noise, A is the system matrix, and B is the
The rate of reaction per volume is usually expressed using input matrix. For the filter implementation, both x̂ and
Arrhenius equation [2]. P must be integrated through time.
E
rX  k o exp C X (2) III. METHOD
 RT 
where k o = reaction rate constant, E = activation energy, A. State Space Representation
R = ideal gas constant, and T = reactor absolute temperature The state space representation of the reactor mass balance
in Kelvin. equation of a CSTR process is presented assuming a linear
Rearranging Eq. (2) and taking the Laplace transform continuous time process with input and measurement noise as
gives the transfer function as in Eq. (3): shown in Fig. 3.
C X ( s) Q
 (3)
C X 0 ( s) E
Vs  Q  k o exp 
 RT 
Table 1 shows the parameters and values used for
simulation in this paper. Fig. 3. Linear continuous-time process with input and output noise.

98

Eze Paulinus C., Muoghalu Chidiebere, Jonathan Amabikutol E., Aigbodioh Ferdinard A., Okoronkwo Emmanuel A., and Ekengwu
Bonaventure O., “Implementation of a Kalman-Bucy filter for estimating product concentration of a CSTR process,” International Research
Journal of Advanced Engineering and Science, Volume 3, Issue 1, pp. 97-100, 2018.
 International Research Journal of Advanced Engineering and Science
ISSN (Online): 2455-9024

x  Ax  Bu  w (8) 1.4
Input

y  Cx (9) 1.2
Estimated Output
Measured Output

ˆy  y  v (10) 1

where u is the vector of inputs, x is the actual states vector,

Product concentration
0.8

y is the actual process outputs vector, ŷ is a vector of the 0.6

actual process outputs, w and v are state and output noise

0.4
respectively. In this paper, the state or process and output
noise are assumed to be zero Gaussian. 0.2

The process, Eq. (4), considered in this paper is a first 0

0 2 4 6 8 10 12 14 16 18 20

order system given by the following state space equation: Time (seconds)

x   x  u (11) Fig. 6. Step Response,  2 =0.01.

yx (12)
where A  1 , B  1 , and C  1
The complete programme for the computer simulation in
Matlab/Simulink environment is shown in Fig. 4. It should be
noted that only the estimated output result is presented in this
paper.

Fig. 7. Step Response,  2 =0.001.

1.4
Input
Estimated Ouput
1.2
Measured Output

1
Product Concentration

Fig. 4. Complete programme for the simulation. 0.8

IV. SIMULATION RESULTS AND DISCUSSION 0.6

A. Simulation Results
Simulations are performed considering different selected
or tuned values for the variance  2 of the process noise and
the variance of the measurement noise. The values were
selected within the range of 0.0001 to 0.1. The simulation
results obtained are shown in Fig. 5, 6, 7, and 8.
Fig. 5. Step Response,  = 0.1.
2
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
Time (seconds)
Fig. 8. Step Response  2
=0.0001.
B. Discussion
In this work, two sources of zero-mean Gaussian noise are
made to enter the system. The input is corrupted with process
noise and the output corrupted with measurement noise. The
variance of the process noise and that of the measurement
noise were tuned within the range of 0.0001 to 0.1. In Fig. 5,
the variance of the process noise and measurement noise were
tuned to 0.1. It can be seen that the estimated signal tracks the
referenced input signal but is highly corrupted by noise. In
Fig. 6, the variance of the process noise and measurement
noise were tuned to 0.01. It can be seen that the estimated
signal tracks the referenced input and is less corrupted by
noise than the result of Fig. 5. In Fig. 7, the process noise and
the measurement noise have their variance tuned to 0.001. The
plot shows that the estimated noise tracks the referenced input
with improved noise rejection. In Fig. 8, the estimated signal
tracks the referenced input with well improved noise rejection.
This shows that by tuning the variance of the process noise
and measurement noise, within the range considered, the

99

Eze Paulinus C., Muoghalu Chidiebere, Jonathan Amabikutol E., Aigbodioh Ferdinard A., Okoronkwo Emmanuel A., and Ekengwu
Bonaventure O., “Implementation of a Kalman-Bucy filter for estimating product concentration of a CSTR process,” International Research
Journal of Advanced Engineering and Science, Volume 3, Issue 1, pp. 97-100, 2018.
 International Research Journal of Advanced Engineering and Science
ISSN (Online): 2455-9024

product concentration of a continuous stirred tank reactor REFERENCES

(CSTR) system can be estimated using Kalman-Bucy filter.
V. CONCLUSION
This work has presented the implementation of a Kalman-
Bucy filter for estimating product concentration in Simulink.
The work considered a typical continuous stirred tank reactor
(CSTR). Assuming ideal operation, the chemical reaction
process taking place in the system was considered as a linear
continuous time process. A model of a reactor mass balance
was used in this work to obtain a transfer function for the
product concentration which is to be estimated using
measurement noise. A Kalman-Bucy filter algorithm was
developed and implemented in Matlab/Simulink environment.
A computer programme using Matlab code for the Kalman-
Bucy filter implemented in Matlab/Simulink embedded block
function was used for the simulation in this paper.
The input and the output were corrupted with zero-mean
Gaussian noise whose variance is tuned within the range of
0.0001 to 0.1. The performance of the designed filter was
presented in terms of the input signal, estimated output and
measured output as shown in the plots in Fig. 5, 6, 7, and 8. It
can be seen from the plots that the performance of the filter in
rejecting the noise and estimating a signal that accurately
tracks the input improves as the value of the noise variance is
tuned within the range of 0.1 to 0.0001.
[1] Persistence Market research, “Continuous flow stirred tank reactor
(CSTR) market: Global industry analysis and forecast 2016-2026”
https://www.persistencemarketresearch.com/market-
research/continuous-flow-stirred-tank-reactor-market.asp
[2] Anonymous, “Mathematical modeling of a continuous stirred tank
reactor,”
[3] Goddard Consulting, “A simple Kalman-Bucy filter in simulink”
hppt://www.goddardconsulting.ca/simulink-kalman-bucy-filter
[4] M. Danish, M. K. Al Mesfer, and M. M. Rashid, “Effect of operating
conditions on CSTR Performance: an experimental study,” Int. Journal
of Engineering Research and Applications, vol. 5, issue 2, pp. 74-78,
2015.
[5] P. K. Igbokwe, J. T. Nwabanne, and S. W. Gadzama, “Characterization
of a 5 litre continuous stirred tank reactor,” World Journal of
Engineering and Technology, vol. 3, issue 2, pp. 25-40, 2015.
[6] R. Antonelli and A. Astolfi, “Continuous stirred tank reactors: Easy to
stabilize,” Automatica, vol. 39, issue 10, pp. 1817-1827, 2003
[7] K. Prabhu and V. M. Bhaskaran, “Controller design for continuous
stirred tank reactor using adaptive control,” Research Journal of Applied
Sciences, vol. 9, issue 8, pp. 489-495, 2014.
[8] M. P. Di Ciccio, M. Bottini, and P. Pepe, “Digital control of a
continuous stirred tank reactor,” Hindawi Publishing Corporation,
Mathematical Probles in Engineering, vol. 2011, Article ID 439785, pp.
1-18, 2011.
[9] R. Alejandro, Rosario C. P. and Juan B. G., “Problem based learning
(PBL): analysisof continuous stirred tank chemical reactors with a
process control approach,” International Journal of Software
Engineering and Application (IJSEA), vol. 1, no. 4, pp. 54-73, 2010.
[10] M. Pottman and D. E. Seborg, “Identification of non-linear process
using reciprocal multiquadratic functions,” J. Process Control, vol. 2,
issue 4, pp. 189-203, 1992.

100

Eze Paulinus C., Muoghalu Chidiebere, Jonathan Amabikutol E., Aigbodioh Ferdinard A., Okoronkwo Emmanuel A., and Ekengwu
Bonaventure O., “Implementation of a Kalman-Bucy filter for estimating product concentration of a CSTR process,” International Research
Journal of Advanced Engineering and Science, Volume 3, Issue 1, pp. 97-100, 2018.